AIRPORT PASSENGER DEMAND FORECASTING USING RADIAL BASIS FUNCTION NEURAL NETWORKS: JUANDA INTERNATIONAL AIRPORT CASE

TITO MUBARAK, Prof. Dr-Ing. Ir. Ahmad Munawar, M.Sc.

2015 | Tesis | S2 Sistem dan Teknik Transportasi

Forecasting merupakan tahap penting dalam perencanaan dan manajemen bandara. Forecasting merupakan alat untuk menentukan apakah terminal baru dan landasan diperlukan dan juga seberapa besar, seberapa banyak dan kapan diperlukan. Namun, seringkali forecasting sering salah dalam memprediksi karena adanya ketidakpastian. Maka dari itu, menangani ketidakpastian merupakan tantangan terbesar dalam forecasting. Forecasting juga merupakan proses mengubah ketidakpastian menjadi risiko yang terukur. Dari hal tersebut, penelitian ini dilakukan untuk mencari metode baru dalam memprediksi penumpang bandara. Sebuah metode bernama Radial Basis Function Neural Networks diperkenalkan untuk memprediksi kebutuhan volume penumpang bandara Juanda. Bandara ini mempunyai posisi penting di Indonesia karena merupakan bandara terbesar kedua dalam hal jumlah penumpang tahunan. Sebagai tambahan, trafik tahunan bandara ini telah melebihi kapasitasnya sebesar 42% pada tahun 2013. Penelitian ini diawali dengan mempelajari pustaka tentang artificial neural networks, time series dan causal forecasting, manajemen bandara dan ekonomi makro. Tinjauan pustaka tentang artificial neural networks, time series dan causal forecasting diperlukan untuk menyusun bagaimana proses forecasting dilakukan. Manajemen bandara dan ekonomi makro dibutuhkan untuk menganalisa secara menyeluruh hasil dari forecasting. Data sekunder diperoleh dari beberapa sumber seperti PT. Angkasa Pura I, Kementerian Perhubungan, Biro Pusat Statistik, Bank Indonesia dan Bank Dunia (World Bank). Tingkat kesalahan secara keseluruhan berada dikisaran 0.2% hingga 16% selama proses training dan testing. Hasil optimal diperoleh dengan tingkat kesalahan di bawah 1% yang dicapai oleh Radial Basis Function Neural Network. Hal ini menunjukkan bahwa metode ini direkomendasikan untuk diaplikasikan di bandara Juanda. Analisa menunjukkan bahwa tambahan kapasitas sangat diperlukan untuk memenuhi kebutuhan bandara saat ini dan di masa depan. Peningkatan kapasitas secara bertahap disarankan hingga mencapai 24 juta penumpang per tahun dalam lima tahun kedepan. Di sisi lain, kejenuhan perekonomian Indonesia yang ditunjukkan oleh Gross Domestic Product (GDP) memberikan risiko yang besar terhadap pengembangan bandara. Jika tidak ada perbaikan situasi ini, maka kekurangan infrastruktur bandara akan lebih parah secara nasional dan khususnya bandara Juanda.

Forecasting is a tool to determine not only whether a new terminal building and runways are required but also how big, how many and when they are required. However, forecasting often faulty because of uncertainties. Since then, addressing uncertainties becomes the greatest challenge in forecasting. Moreover, forecasting is a part of process of converting uncertainties into qualified risks. Therefore, this study was held to search a new method in airport passenger forecasting. A method namely Radial Basis Function Neural Network is introduced as an airport passenger demand forecasting tool in the Juanda Airport. The airport has important role since it is the second largest airport in term of annual passenger in Indonesia. In addition, the current annual passenger traffic of the airport has surpassed its basic capacity by 42 % in 2013. The research was initiated with an extensive literature study on artificial neural networks, time series and causal forecasting, airport master planning, and macro economy. The literature study of artificial neural networks, time series and causal forecasting are needed to construct how the process of forecasting has to be developed. The airport master planning and macro economy study are needed in order to analyze entirely the forecasting results. Secondary data are collected from several sources such as PT. Angkasa Pura I, Indonesia Ministry of Transport, Biro Pusat Statistik, Bank Indonesia and The World Bank. The mean squared error (MSE) ranges between 0.2% and 16% during training and testing processes. The optimum result shows that the Radial Basis Function Neural Networks gained error below 1%. It means that this method is highly recommended to be utilized in the Juanda Airport. The analyses show that additional capacity is urgently needed to satisfy the current and future demand. The analyses also suggest that the capacity has to be expanded gradually near to 24 million passengers in the next five years. In the meantime, the saturation of the Indonesian economy which is reflected by Gross Domestic Product (GDP) provides high risks for further airport expansion. If there is no improvement in the macro economy of Indonesia, the airport capacity bottlenecks seem to be worsening nationwide and particularly in the Juanda Airport.

Kata Kunci : forecasting, artificial neural networks, Radial Basis Function, transportasi udara, bandara, bandara Juanda